openwashdata · larnsce · Jul 12, 2026 · Jul 12, 2026
diff --git a/CLAUDE.md b/CLAUDE.md
@@ -70,6 +70,7 @@ This is the openwashdata newsletter repository for managing monthly newsletter c
 ## Project Management with GitHub CLI
 
 - Create branches from `dev`. Do not push to `main`
+- Always open pull requests against `dev`, never against `main`
 - List issues: `gh issue list`
 - View issue details: `gh issue view 80` (e.g., for issue #80 "Rename geographies parameter")
 - Create branch for issue: `gh issue develop 80`

diff --git a/strategy-growth.md b/strategy-growth.md
@@ -0,0 +1,242 @@
+# Newsletter Strategy (Phase 2): Engagement & Growth to 1,000
+
+Goal: grow the newsletter to 1,000 subscribers within 12 months and
+make each issue something readers act on (reply, click, share, submit),
+not just read.
+
+This strategy replaces the Phase 1 focus ("inform the community about
+our infrastructure") with a reader-first focus: every issue must help a
+WASH professional do something with data.
+
+## 1. Where we are: audit of issues 1-21
+
+Findings from reviewing every published issue (2023-08 to 2026-03):
+
+### What works
+
+- The course is the strongest asset. ds4owd-001 drew 200+
+  registrations from 80+ countries; the special issue (#19) promoting
+  ds4owd-002 was the only issue written entirely around reader benefit,
+  and it is the best issue in the archive.
+- Events create spikes. The PLOS webinar (39 participants), two
+  hackathons, SWWW and UNC sessions all generated content and contact
+  with new people.
+- Contributor of the Month is a genuine differentiator. It puts
+  community members' names, faces, and links in front of peers.
+- Credibility markers exist: 29+ published datasets, CRAN package
+  (`washr`), Swiss National ORD Prize 2024, ETH backing, partnerships
+  with WASHWeb, SOIL, PLOS, and BASEFlow Malawi.
+
+### What is broken
+
+1. Cadence collapsed. Monthly held until mid-2024, then: issue 18
+  opens with "It's been more than four months since our last
+  newsletter", issue 20 is Sept 2025, issue 21 is March 2026. A list
+  that hears from you 3-4 times a year cannot compound growth, and
+  irregular sending trains readers to ignore or unsubscribe.
+2. Content is producer-centric. Most issues are "we released X, we
+  attended Y, we hired Z". Dataset announcements are lists of package
+  names with no chart, no finding, no reason for a non-R-user to care.
+3. Zero interaction mechanisms shipped. The Phase 1 plan included a
+  footer poll ("What's your favorite section?"). It was never
+  implemented. No issue asks a question, invites a reply, or offers
+  anything to click other than outbound links.
+4. Quality control is missing. Shipped errors include: issue 4 swaps
+  the `waterpumpkwale` and `grdwtrsmpkwale` URLs; issue 13 links
+  `portawaterperu` to the saniabidjan site; issue 14's headline says
+  "Three new datasets... Uganda, Malawi, Brazil, Cote d'Ivoire, Peru"
+  above two datasets from Ghana and a global source, and links
+  `whatawaste` to the waschoolpiracema site; issue 17 has two dangling
+  sentences ("apply:" and "accessible here:" with no link).
+5. No measurement loop. Analytics were "TODO" in Phase 1 and remain
+  unreviewed. We do not know our open rate, click rate, or list size
+  trend, so nothing gets tested or improved.
+
+## 2. Goals and targets
+
+Baseline: pull the current subscriber count, open rate, and click rate
+from Buttondown in week 1 and record them in this file. Targets below
+assume a starting list of roughly 200-400.
+
+| Metric                    | Now       | +6 months | +12 months |
+|---------------------------|-----------|-----------|------------|
+| Subscribers               | (record)  | 550       | 1,000      |
+| Issues sent on schedule   | 3-4/year  | 6/6       | 12/12      |
+| Open rate                 | (record)  | 45%       | 50%        |
+| Click rate                | (record)  | 8%        | 12%        |
+| Replies + poll votes      | ~0/issue  | 10/issue  | 25/issue   |
+| Reader submissions        | 0         | 1/issue   | 2/issue    |
+
+Net growth needed: ~600-800 subscribers, i.e. ~55-65/month average.
+This is not achievable through organic forwarding alone; it requires
+the acquisition channels in section 5, with the course as the anchor.
+
+## 3. Positioning: who is the reader
+
+Write for the WASH professional or student who works with data but is
+not (yet) part of our community: an M&E officer at an NGO, a
+researcher, a utility analyst, a course alumnus. They ask:
+
+- "Is there data I can use for my region/topic?"
+- "How do I do this analysis / make this map / share my data?"
+- "What are peers doing that I can learn from?"
+
+Editorial test for every section: does this answer one of those three
+questions? Team news, funding updates, and internal milestones may
+appear, but never lead, and never fill an issue.
+
+## 4. Content system
+
+Fixed monthly skeleton. Short beats late: a 3-section issue on
+schedule beats a 6-section issue two months later.
+
+1. One finding, one chart (lead section). Take ONE dataset (new or
+  from the back catalogue of 29+) and show a single finding with one
+  visualization and 3-4 sentences of interpretation. Link to the
+  package and to a short "how this chart was made" post. This turns
+  dataset announcements from lists into stories, and the back
+  catalogue means content never depends on new releases.
+2. New releases (compact). Bullet list of new packages, one line
+  each, links verified. No release this month? Skip the section.
+3. Practical tip. One task-oriented snippet: a map recipe, a data
+  dictionary trick, an AI-assisted cleaning workflow, a `fairenough`
+  or `washr` feature. Reuse and repackage course material — it is
+  already written and proven to attract this audience.
+4. Community corner (interaction hub, see section 5). Contributor of
+  the Month, reader showcase, poll results from last issue, this
+  issue's question.
+5. Get involved (footer). Keep the current compact version.
+
+Recurring annual content: course launch special (the #19 format,
+proven), hackathon showcase, end-of-year wrap-up with community stats,
+graduate capstone showcases.
+
+Writing rules: per CLAUDE.md (no "thrilled/excited/remarkable", plain
+factual language). Subject lines state the concrete payoff ("How water
+costs vary across Accra households"), not "Newsletter #22". Preview
+text is written, not left to default.
+
+## 5. Interaction plan
+
+Every issue must contain exactly one primary question and one reply
+invitation. More than that dilutes response.
+
+- Poll in every footer. Buttondown supports click-tracking links;
+  a 3-option poll ("Which section was most useful?") costs one click.
+  Publish results in the next issue so voters see their input used.
+- Monthly question. One concrete question in the community corner,
+  e.g. "What dataset do you wish existed?" or "What's blocking you
+  from sharing your data?". Answers via reply. Best answers quoted
+  (with permission) in the next issue.
+- Reader showcase. Standing call: "Made something with openwashdata
+  packages? Reply with a link." Feature one per issue with name,
+  chart, and links — same mechanic as Contributor of the Month, but
+  open to any reader, which gives every subscriber a path to being
+  featured.
+- Reply-to is a monitored human mailbox, and every reply gets a
+  personal answer within a week. Nothing kills interaction faster
+  than replies going into a void.
+- Data challenge (quarterly). Small task on one dataset ("make one
+  chart from `washmalawi`"), submissions via reply or chat, winners
+  featured. This is the hackathon mechanic, made remote and async.
+- Annual reader survey (5 questions max), results published.
+
+Interaction is also a retention lever: people who have replied, voted,
+or been featured do not unsubscribe.
+
+## 6. Growth plan: where 600-800 subscribers come from
+
+Ranked by expected yield:
+
+1. Course pipeline (~300-400). Add a newsletter opt-in checkbox to
+  every ds4owd registration form, opt-in at certificate delivery, and
+  a sign-up link in course materials and lecture closings. Cohort
+  registration volumes (200-300+) make this the single largest
+  channel. Run one cohort launch special issue per cohort (the #19
+  format) and ask registrants to forward it.
+2. Dataset package websites (~100-150). 29+ package sites are our
+  SEO surface — people find `gdho` or `glaas` via search. Add a
+  one-line subscribe link to the `washr`/`fairenough` website
+  template footer so every current and future package page carries
+  it. Same for the openwashdata.org blog, data gallery, and archive.
+3. Events (~50-100). Every webinar, hackathon, SWWW/UNC session ends
+  with one slide: QR code to subscribe. Follow up with attendee lists
+  where consent allows. Target 3-4 such events in 12 months —
+  webinars are the cheapest to run and drew 39 sign-ups with minimal
+  promotion last time.
+4. Partner cross-promotion (~50-100). WASHWeb, SOIL, BASEFlow, PLOS,
+  and GHE/ETH channels each mention the newsletter once or twice a
+  year (newsletter swap, blog mention, or social post). One
+  coordinated ask per partner, reciprocated.
+5. Referrals (~50). "Forward this email" is currently a dead line in
+  the footer. Make it an explicit campaign twice a year: "Know one
+  colleague who works with WASH data? Forward this issue." Buttondown
+  subscriber counts by referral source tell us if it works.
+6. Team social (~50). Each issue gets one LinkedIn post from the team
+  highlighting the lead finding (the chart travels well), linking to
+  the web archive version with a subscribe CTA.
+
+Every channel must land on a single canonical sign-up page with a
+one-sentence value proposition and a recent issue preview.
+
+## 7. Cadence and workflow
+
+- Monthly, first Monday, 12 issues in 12 months, no skips. If a month
+  is thin, send the short version (lead finding + community corner +
+  footer). The special issue (#19) proves short works.
+- The monthly cycle from Phase 1 stays, with one change: Week 2 now
+  includes a 30-minute metrics review (section 8) and logging replies
+  and poll results for the next issue.
+- Ownership: name one editor per issue (rotating is fine) who is
+  accountable for the send date. The 4-6 month gaps happened because
+  the newsletter was everyone's job and no one's deadline.
+- Draft in `.qmd` with R chunks for package metadata (current
+  practice, keep), render to `.md`, archive after send.
+
+## 8. Quality checklist (before every send)
+
+- [ ] Every link opened and verified (issues 4, 13, 14 shipped wrong
+      package URLs; this is a recurring failure mode)
+- [ ] Package names match their URLs
+- [ ] Headline counts/countries match the section content
+- [ ] No placeholder text or dangling sentences (issue 17 shipped two)
+- [ ] One question + one poll present, reply-to verified
+- [ ] Subject line and preview text written
+- [ ] Test send read on mobile
+- [ ] Alt text on images
+
+## 9. Measurement
+
+Monthly (Week 2, 30 minutes, log in this repo):
+
+- Subscriber count, net growth, growth by source
+- Open rate, click rate, top 3 clicked links
+- Replies and poll votes received
+- One decision recorded: what to change next issue
+
+Quarterly: review against the section 2 table. If growth is under 50%
+of target for two consecutive quarters, the course pipeline (channel
+1) is underperforming and needs direct attention first — it is the
+only channel large enough to close the gap.
+
+Tools: Buttondown analytics (already available), Plausible for
+openwashdata.org referral tracking, UTM parameters on all newsletter
+links so we can see what newsletter traffic does on the site.
+
+## 10. 12-month roadmap
+
+- Month 1: Record baseline metrics. Set up canonical sign-up page,
+  poll links, QA checklist. Resume monthly sending with the new
+  skeleton. Add opt-in to course forms.
+- Month 2: Add subscribe links to package site template and blog.
+  First poll results published. First LinkedIn post per issue.
+- Month 3: First quarterly data challenge. Partner cross-promotion
+  ask #1 (WASHWeb).
+- Months 4-6: First referral campaign. One webinar with subscribe
+  CTA. Quarterly review #1.
+- Months 7-9: Course cohort launch special + registration opt-in
+  harvest. Second data challenge. Partner asks #2-3 (SOIL, BASEFlow,
+  PLOS). Quarterly review #2.
+- Months 10-12: Reader survey. Second referral campaign. Second
+  webinar or hackathon showcase. End-of-year wrap-up issue with
+  community stats. Final review against 1,000 target.
diff --git a/strategy.md b/strategy.md
@@ -1,5 +1,9 @@
 # Newsletter Strategy (Phase 1)
 
+> **Note**: This document is superseded by
+> [strategy-growth.md](strategy-growth.md) (Phase 2: engagement and
+> growth to 1,000 subscribers). It is kept for reference.
+
 The first phase of the newsletter focuses on informing the community of our infrastruture and long-term goals.
 
 **Objectives:**